Click for next page ( 54


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 53
5 Reproductive Patterns and Children's Health - Despite the major improvements in child health that have occurred since World War II, infant and child mortality rates in many developing countries remain very high. During 1980-1985, almost 90 out of every 1,000 infants born in the developing world died before their first birthday. In contrast, there were an estimated 16 infant deaths per 1,000 births in the developed world (United Nations, 1988b). Other indicators of poor child health, such as the incidence of infectious disease and malnutrition, also remain high in many developing coun- ~ies, particularly in the poorer countries in sub-Saharan Africa and South Asia. The major difference in child health between developing and industrialized countries is that infectious and parasitic diseases, including diarrhea! diseases, and malnutrition, are considerably more common in the Third World. In addi- tion, children in developing countries are likely to have multiple conditions that increase the potential severity of illness and raise the probability of death. To illustrate differences in the distribution of causes of death in countries with different patterns of mortality, Table 5.1 shows the distributions of cause of death for infants in a high-mortality population (Recife, Brazil, in the late 1960s), a moderate-mortality population (Paraguay in 1983), and a low-mortality popula- tion (the United States in 1983~. Infectious, parasitic, and respiratory diseases, like measles and diarrhea! disease, account for almost two-thirds of the infant deaths in Recife, but only 6 percent of the infant deaths in the United States. Because deaths due to infectious diseases become much less common as mortal- ity declines, congenital anomalies and conditions associated with birth and the immediate postbirth period are relatively more prominent in low-mortalibr popu- lations. 53

OCR for page 53
54 CONTRACEPTION ID REPRODUCTION TABLE 5.1 Cause of Death (Percentage of All Infant Deaths) and Infant Mortality Rate for Free Populations Cause of Death Recife, Brazil 1968-1981@ High Mortality Paraguay United States 1983 1983 Moderate Monality Low Mortality Infectious and parasitic 51 23 diseases Diseases of respiratory 11 14 system Congenital anomalies 4 4 Certain pennatal 26 24 conditions Ill-de6~ned All other 8 21 Infant mortality rate 91 S1 13 2 4 21 47 14 11 11 a"Basic causes" only, by ICDD-8 asssification (from ~ter-Arnerican Investigation of Mortality in Childhood). Sources: Puffer and Serrano (1973); World Health Organization (1987a). Causes of poor health and mortality also change during the course of child- hood. Congenital problems, low birthweight, and difficulties during pregnancy or birth are more likely to affect morbidity and mortality during the neonatal period (the first month of life) than later in infancy. Certain infections, such as neonatal tetanus, are also particularly prevalent during the first month, and other infectious diseases, such as pneumonia, are significant neonatal health risks. Mortality and illness after the neonatal period, which are usually associated with infectious or parasitic diseases and poor nutrition, are more directly influenced by the environment in which a child lives than mortality and illness during the first month of life. Specific infectious and parasitic diseases often affect children of particular ages. For example, Foster (1984) reports differential effects of measles by age, with highest mortality under age 1 and decreasing mortality thereafter. The age pattern of childhood illness in high-mortality countries often depends on breastfeed- ing and weaning practices. Once weaning begins, children lose the immunities provided by breast milk, and they begin to consume food that may be contami-

OCR for page 53
REPRODUCTIVE PANELS ED CNI=REN'S HEARTH 55 nated. After weaning, children become dependent on the family food supply, which may be inadequate, contaminated, or inappropriate for their needs. In addition, protection from maternal antibodies declines with age. Because the etiology of poor health and mortality can change considerably during infancy and childhood, it is important to consider the effect of reproductive patterns on child health separately for children of different age groups. Previous research suggests that the risk of mortality and poor health are higher for children who are born to mothers with particular reproductive histories. Results of bivariate studies) usually show that infant and child~mortality rates are higher for those who are: the firstborn, born to a young mother, or a combination; a higher-order birth, born to an older mother, or a combination; born into a family with a large final number of children ever born; born before or after a short interbirth interval. It has also been hypothesized that children born as a result of unwanted pregnancies are likely to be in poorer . . . health compared with children born as a result of other pregnancies. In this chapter we fast review sources of data and analytical issues and then examine studies on the relationships between reproductive patterns and infant and child health. These studies are summarized at the end of the chapter in Appendix Table S.A. SOURCES OF EVIDENCE Much of the earliest evidence linking child health and survival, particularly during infancy, with maternal age, birth order, and the timing and spacing of births was based on data from industrialized countries. Many of these studies dealt with small, select populations and had few, if any, controls for confounding factors. Recently, a number of population-based studies using more sophisticated statistical procedures and data from developing countries have provided substan- tially more information about these complex relationships between reproduction and health in Third World settings. Most of these studies focus on infant and child mortality as measures of health, because data on other indices of health are not as commonly available. For this reason, the results described in this chapter draw primarily on studies of the associations between reproductive patterns and child survival. Whenever possible, however, the discussion is supplemented with other information about child health, such as birthweight and illness. Much of the recent evidence on the association between reproductive patterns and child health in developing countries comes from the World Fertility Survey ~See, for example, Rutstein (1983) for evidence on the first four types of characteristics listed.

OCR for page 53
56 CONTRACEPTION AND REPRODUCTION (WFS), which were conducted between 1974 and 1982 in a number of developing countries. In these surveys, nationally representative samples of women of reproductive age were interviewed about their birth or pregnancy histories and the fate of each of their live-born children. In analyses of these data, this birth history information has been used to determine the length of the intervals preceding and following the birth of a given child, the child's birth order among his siblings, and his mother's age at the time he or she was born. Recent analyses of data from the WFS and other retrospective survey data have demonstrated that child survival is strongly associated with longer intervals between births. This association has been found in a large number of populations with very different levels of mortality, fertility, and economic development. These studies have also shown that maternal age and birth order are significantly related to child survival in many populations. However, Here are several poten- tial problems with drawing inferences about causal relationships from analyses of these data. First, reporting errors common to retrospective fertility histories may exaggerate the relationships observed between birth spacing and child survival.2 Second, most of the surveys contain only limited information on breastfeeding, length of gestation, birthweight, and other biomedical characteristics of the mother and child that may be important factors in the relationship between child health and reproductive patterns. Third, many of the surveys collected only limited information on socioeconomic status and other family characteristics, which may independently affect both fertility and children's health. Fourth, data imputation procedures used in the WFS may alter estimates of effects Russell and Ro- driguez, 1989~. Because of these limitations on analyses of retrospective survey data, we also draw on the results of studies based on two other types of information. One source is a small number of studies based on data collected longitudinally in developing countries. Carefully collected longitudinal data do not suffer from the same We of systematic misreporting that affects retrospective surveys. Further- more, longitudinal studies frequently also collect detailed data on biomedical and behavioral factors, such as length of gestation and birthweight, which may not be available in retrospective data. Unfortunately, longitudinal data have been col- lected in only a few populations, and, in some cases, more extensive analyses of extant data needs to be undertaken. Another source of evidence is studies of historical populations in Europe and the United States that experienced mortality rates at the same or higher levels as those found in contemporary populations in the Third World. One advantage of historical data is that, like longitudinal data, they are less subject to the type of misreporting problems sometimes found in retrospective survey data Historical 2 previous research has shown that respondents are more likely to anit both the birth and death of children no longer alive at the time of the interview. bile women who do anit reports of children who have died will appear to have both longer birth integrals and their children will have lower mortality rates (Power, 1988; Oeland and Sathar, 1984).

OCR for page 53
REPRODU=WE PA=E~S ID CHI=REN'S HEALTH 57 data sets, however, suffer from different types of data quality problems, including omission of events due to immigration and lapses in recordkeeping (Lynch, 1987; Bean et al., 1987). With historical data, reproductive and mortality histories cover the entire reproductive span. Comparison of results from high-mortality historical populations with those from contemporary developing countries also allows us to make inferences about whether the hypothesized relationships are common to high-mortality populations in a variety of cultural, social, and eco- nomic contexts. Assessment of the association between reproductive variables and child sur- vival is complicated by a number of methodological, statistical, and theoretical problems (see Potter, 1988; Rosenzweig and Schultz, 1983; Hobcraft et al., 1985; Hobcraft, 1987; Pebley and Millman, 1986~. First, the reproductive variables of interest are likely to be highly correlated. For example, higher child mortality observed among children born to teenage mothers may actually be a consequence of the fact that a large proportion of these births are first births. Unless both variables are included simultaneously in the analysis, higher risks of mortality may be incorrectly attributed to either one variable or the other. Second, reverse or spurious causality complicates the interpretation of results unless adequate statistical controls are introduced for factors such as breastfeed- ing and the survival status of the preceding birth. For example, an apparent association between a child's mortality risk and the length of the subsequent birth interval could be due to either the child's death ending breastfeeding and leading to earlier conception of the next child, or to early weaning or no breastfeeding, which itself places the child at greater risk. Third, certain characteristics of the family may increase the likelihood both that children in that family will be in good health and that births will be widely spaced, or that there will be a small number of children in the family. For example, women who have completed elementary education may be both more likely to use contraception to space or limit their births and to be able to provide better care for their children. In addition, in a noncontraceptive population, women who lose children will have more total births because of shorter birth intervals caused by interrupted lactation. If adequate attention is not paid to the role of unobserved heterogeneity in the design of the analysis, reproductive variables will appear to be correlated with child health, when in fact the relation- ship is not causal.3 Because fertility and child health are both affected by parental choices, unobserved variable~environmental constraints, biologically fixed characteristics, or parent preferences can affect both outcomes, and any asso- ciation between fertility and child health may be a biased estimate of the causal relationship. There is no consensus about the most appropriate way to eliminate potential bias. 3 This problem can also be viewed as simultaneous-equations bias, the consequences of which are well described in the economics literature (see, for example, Schultz, 1984).

OCR for page 53
5 ~ CONTRACTION kD REPRODUCTION Many recent studies of reproductive patterns and child health have used multivariate statistical methods in an attempt to control for some of these prob- lems. Although limitations remain in the studies on which our assessment is based, we rely principally on analyses that employed multivariate methods and introduced statistical controls for potentially confounding factors. Both the theoretical models and statistical methods used in this research continue to evolve. EFFECTS OF BEING FIRSTBORN AND YOUNG MATERNAL AGE Children born to teenage mothers and children who are firstborn are generally at higher risk of dying than children born to mothers in their twenties and children of birth order 2, 3, and 4. In several studies, both young maternal age and first birth order remained important predictors of infant and child mortality, even when the other variable is held constant. These results indicate that the estimated effect of young maternal age is not due solely to the fact that births to young mothers are more lilcely to be fast births, nor is the estimated effect of being firstborn due solely to the fact of being more likely to have a young mother. First Births The available evidence from many countries suggests that the negative effects of being firstborn may be limited to the first year of life. In an analysis of World Fertility Survey data from 34 countries, Hobcraft (1987) found that the average estimated risk of dying across all countries for firstborn children compared with children of birth order 2 and 3 with optimal spacing was 1.7 for the neonatal period, 1.5 for the postneonatal period. No excessive risk was found for the toddler period (ages 1 to 2) and for childhood (ages 2 to 5).4 It is important to note, however, that there is considerable variation in the size of the relative risks for first births among the national populations included in the Hobcraft analysis and among the populations of other studies that have looked at the same issue. Indeed, in five countries in the Hobcraft analysis, firstborn children do not experience higher risks of death. The results from other multivariate analyses are mixed: some find higher risks for Olrst births and others do not. We have examined variations in the sizes of the relative risks for fast births across coun- tnes and found no systematic relationship between the relative risks and either total fertility rates or infant mortality rates. 4 The figures given in the text are the relative odds of dying for firstborn children compared with children who were of second or third births with favorable birth spacing (more than 24 months) and with no prior child deaths. Hobcraft (1987) points out that the effects of being a first both may be somewhat exaggerated by comparison only with second and third births, when there were no previous child deaths (see Hobcraft, 1987, pg. 8, for further discussion).

OCR for page 53
REPRODUCTIVE PATTERNS AND CHI=REN'S HEALTH 59 Hypotheses about the higher risks of mortality and poor health associated with first births usually center around physiological adjustment of the mother to her fast pregnancy. There is considerable evidence to indicate that first pregnancies and births have a higher rate of complications than subsequent pregnancies. For example, Fortney et al. (1985) present evidence from 86 hospitals, mostly in developing countries, indicating that, although the incidence of complications of delivery, such as breech presentation, is not higher for first births, pennatal mortality rates associated with such complicated deliveries of first birds are higher. Other complications of pregnancy, such as pregnancy-induced hyperten- sion, appear to be more common for women who are pregnant for the first time (Haaga, 1989~. These complications of pregnancy and childbirth result in both a higher incidence of maternal morbidity and mortality and in a greater risk of morbidity and mortality for infants. There is also evidence that the incidence of low birthweight (less than 2,500 g) is higher among first births (DaVanzo et al., 1984; Niswander and Gordon, 1972~. Recent reviews by Haaga (1989) and Kramer (1987) suggest that much of this higher incidence of low birthweight among firstborn children is due to intrauter- ine growth retardation rather than to prematurity. Haaga (1989) suggests that the higher incidence of malaria infestation in some areas may account for some of the excess health risk to firstborn children during first pregnancies. The presence of malarial parasites in the placenta is associated with low birthweight. Research in sub-Saharan Africa indicates that women having their first pregnancy have twice the rate of placental malaria as women who have already been pregnant before (Bray and Anderson, 1979; McGregor et al., 1983~. However, more evidence is needed before we can assess the role of malaria in differences in infant mortality rates by birth order. Young Maternal Age The Hobcraft (1987) cross-sectional analysis of WFS data indicates that the risks for children of teenagers are higher than for those of older women. On average, across 34 countries, Hobcraft found that the odds of dying for the children of teenage mothers were 1.2 times those for mothers ages 25 to 34 during the neonatal period, 1.4 times during the postneonatal period, 1.6 times during the toddler years, and 1.3 times during childhood years. Again, however, there was substantial variation among countries included in the study in terms of the size of the effects of young maternal age. At least two explanations for the observed association between young mater- nal age and elevated risks of child mortality have been suggested. First, pregnancies that occur before full maternal growth or physical maturation is achieved may place both the woman and her child at greater risk of complications of pregnancy and childbirth. There is some evidence that very young maternal age may have negative consequences for children because of the greater likeli- hood of birth trauma (silken and Walls, 1986~, but Haaga concludes that the

OCR for page 53
60 CONTRACEPTION ED REPRODUCTION evidence that competition for nutrients exists between maternal growth and fetal growth in women who have not yet reached full physical maturity is weak. To study the biological mechanisms involved in the relationship between young maternal age and child health ideally requires analyses that look at the effects of gynecological age (the stage of physical maturation that a girl has achieved), rather than chronological age. This issue may be more salient in developing countries in which the mean age at menarche is relatively late (Foster et al., 1986) and, in some countries, the proportion of girls having their first birth shortly after menarche is considerably higher. Because of the difficulty of assessing gyneco- logical or biological age for large samples, studies on which our conclusions are based rely principally on chronological age. Even if we are limited to chronologi- cal age, it is important to distinguish between very young maternal ages (less than 17), which may be particularly problematic, and the later teenage years (18-19), which may be optimal, at least physiologically, for childbearing. However, most of the research related to maternal age considers all ages less than 20 together. The second possible explanation is that young women who become pregnant are less likely to receive early and adequate prenatal care, more likely to be from poor families, and less able to care for their children because they have not reached full psychological maturity themselves. Recent reviews of evidence linking teenage childbearing to poor child health in the United States (Strobino, 1987; Geronimus, 1987; McAnarney, 1987) have concluded that the main reason for this association is that teenage mothers are more likely to be socially disad- vantaged than women who give birth at older ages. This explanation is probably less applicable in many developing countries, in which births to teenagers are more common and usually take place within marriage or a socially sanctioned union. Furthermore, evidence from several studies in developing countries sug- gests that the association between higher risks of child mortality and young maternal age persists even when socioeconomic status is held constant. However, it may be that the measures of socioeconomic status and living conditions used in these studies do not adequately capture characteristics common to teenage moth- ers in developing countries that affect child survival, such as inadequate use of prenatal care. Summary From the evidence available, we conclude that firstborn children and children born to very young mothers are at greater-than-average risk of mortality and poor health. In the case of firstborn children, there is some evidence to suggest that there are physiological reasons for this greater risk. In particular, women who are pregnant for the fast time are more likely to experience complications of preg- nancy and childbirth, their babies are more likely to be of low birthweight, and, in some areas, maternal malaria, which is more common during first pregnancies, may contribute to the higher mortality of firstborn children.

OCR for page 53
REPRODUCTIVE PATTERNS AND CNi=REN'S HEALTH 61 The issue of age is more complicated. Although some evidence suggests that births to very young girls may be at higher risk for physiological reasons, the effect may be principally social and psychological for older teenagers. Even in countries with relatively high birth rates for girls younger than 17, the proportion of all teenage births that occur to very young teenagers (17 and younger) is relatively small. The evidence from studies that group teenagers of all ages together is mixed. More research is needed in this area. EFFECTS OF HIGH BIRTH ORDER AND OLDER MATERNAL AGE The available evidence suggests that the detrimental effects of high birth order and older maternal age on child survival are not as important as the effects of young maternal age, being firstborn, or close spacing between births. Results of multivariate studies of this issue have produced divergent results. Hobcraft et al. (1985) conclude that much of the elevated risk attributed to older maternal ages and higher birth orders is probably produced by close birth spacing. Although their results vary considerably among the 34 countries studied, on average the risk was higher for parities 7 and higher. Pebley and Stupp (1987) also report higher risks for children of older mothers and higher birth orders in Guatemala, but neither Gubhaju (1986) nor DaVanzo et al. (1983) find significant effects of older maternal age and high birth order on child survival in Nepal and Peninsular Malaysia, respectively. Two studies using historical data (Bean et al., 1987, and Knodel and Hermalin, 1984) did find higher infant mortality at older maternal ages and higher birth orders. However, Knodel and Hermalin show that final sibship size (i.e., the total number of births the mother eventually has) is a more important correlate than birth order, and Bean et al. found that the likelihood that all previous children survived through infancy was more important for survival than birth order for second- and higher-order births. There is less evidence concerning independent effects of older maternal age on infant health than there is on birth order. Older maternal age is associated with an increased incidence of congenital abnormalities, including Down's syndrome (Hansen, 1986; Hook, 1985), but these congenital abnormalities are a relatively minor cause of infant death in developing countries (Haaga, 19893. EFFECTS OF SHORT BIRTH INTERVALS The evidence concerning the effects of birth spacing on child survival and health is more consistent than that concerning the effects of high parity and maternal age. Studies based on very different types of data from culturally and socially diverse populations consistently find a negative association between short birth intervals and a child's chances of survival. This is especially true for the length of the previous interval, i.e., the length of time since the birth of the previous child.

OCR for page 53
62 CONTRACEPTION kD REPRODUCTION Results from several studies indicate that the crucial period is up to 24 months after the birth of the previous child. Children born within this period are at considerably higher risk than children born after longer previous birth intervals. For example, the results from the Hobcraft (1987) study show the average risk of dying for children born within two years of a previous sibling relative to children born after longer intervals is approximately 1.8 in the first year and 1.3 for toddlers (ages 1 to 2) and in the childhood years (ages 2 to 5~. Most studies have found that the mortality risks associated with short intervals are significantly higher when the child who begins the interval dies before the next child is born. This may be due to household or familial effects that increase risk for all children. However, these studies did not control for potentially confounding factors, such as breastleeding, that may also influence the relationship. There is considerably less evidence on the association between length of previous birth interval and other health indices such as birthweight and growth. Low birthweight may be due to intrauterine growth retardation, defined as birth- weight less than the 10th percentile for gestational age, or preterm delivery, defined as gestational age less than 37 weeks.5 Studies of the relationship between birth interval length and birthweight are complicated by the fact that preterm births, by their nature, have shorter periods of gestation and thus shorter birth intervals. These confounding effects associated with preterm births need to be controlled to estimate the association between birth interval length and birth- weights accurately. Although the risk of death is increased for all low-birthweight infants, the risk is highest for preterm babies, especially those with very low birthweights. Both differentiation between intrauterine growth retardation and preterm births and measurement of birth-to-conception intervals are necessary to determine accu- rately possible relationships. Ferraz et al. (1988) report a significantly increased relative risk of intrauterine growth retardation associated with interpregnancy intervals of six months or less. No association was found between birth-to- conception intervals and preterm delivery. Several investigations have shown an association between short intervals between the birth of a child and the concep- tion of the following child and an increased risk of low birthweight, though this was not observed in studies in Norway and the United States (Erickson and Bjerkedal, 1978; Klebanoff, 1988~. Fewer studies have attempted to investigate the association between child survival and the length of the following interval. Excess mortality of those born before a short succeeding interval may be due to early cessation of breastfeeding and resultant inadequate feeding and increased exposure to pathogens at vulner- able ages, but estimation of such effects must also allow for reverse causation. 5 Kramer (1987) examined the extensive literature on low binhweight, concluding that research findings are frequently Conflicting because of a failure to distinguish between intrauterine growth retardation and pr~xnaturity, inadequate control for confounding variables, and lack of statistical power.

OCR for page 53
REPRODUCTIVE PATTERNS AND CHILDREN'S HEALTH 63 The death of the child may itself cause the subsequent interval to be short either through a biological effect (abrupt end of lactation leading to short postpartum amenorrhea) or a behavioral effect Tents trying to replace the lost child quickly). Nonetheless, studies that have examined the issue and controlled for reverse causality have generally found that short subsequent intervals are associated with higher mortality risks for the child whose birth begins the interval. For example, on the average across 34 countries, Hobcraft (1987) found that the risk of dying as a toddler (i.e., in the second year of life) for children whose mother~had another birth within 12 months of their own was 2.2 times the risk of children whose mothers delayed the next birth for at least 18 months. As in the case of other results cited above, it is important to note that the risks associated with being born before a very short interval vary considerably among the countries included in the Hobcraft study and in other studies that have examined this relationship. Possible Factors There are several mechanisms through which short birth spacing may increase a child's risk of dying. Although some evidence is available concerning some of these mechanisms, the information is not sufficient to allow us to say with confidence exactly why close birth spacing is associated with higher child mortal- ity. Furthermore, the relative importance of each mechanism may vary consid- erably among populations. One mechanism through which closely spaced births may affect a child's health is by reducing the time available to the mother to recover from one pregnancy before beginning the next, leading to the birth of a less-than- normally healthy child. Breastfeeding and, to a lesser extent, pregnancy are significant drains on a woman's nutritional resources (Merchant and Martorell, 1988~. There is evidence that short birth and pregnancy intervals are associated with low birthweight (Fedrick and Adelstein, 1973; DaVanzo et al., 1984; Fortney and Higgins, 1984~. However, several studies that have attempted to link indices of maternal health to short birth spacing have not produced persuasive evidence that so-called maternal depletion accounts for the association between birth spacing and child survival except in extremely malnourished populations (Winikoff and Sullivan, 1987; Ferraz et al., 1988; Costello, 1986; Pebley and DaVanzo, 1988~. A; second possible mechanism is that in families with closely spaced births, there may be greater competition among children of approximately the same age for scarce family resources. These resources may include not only food, clothing, and living space but also parental time and attention. Competition between siblings may occur because of either short previous or subsequent intervals. One obvious example of competition occurring because of a short subsequent interval is that the conception of another child often means that the mother weans the child she is breastEeeding. Analyses of the determinants of infant and child mortality that have investigated the effects of breastfeeding have routinely shown

OCR for page 53
REPRODUCTIVE PATTERNS AND CHI=REN'S HEALTH 65 Some characteristics of families that may affect the observed relationship are, in practice, unobservable themselves. Two examples are fecundity, or the ability to become pregnant, and frailty, or the underlying (possible genetic) predisposition toward illness; unobserved heterogeneity of this sort can be dealt with only by the use of experimental designs. Other characteristics include a family's propensity to use health and family planning services and its attitudes and the skills of its members related to planning and intervention in natural processes that are likely to be positively associated with birth spacing in developing countries. In this case, the omission of a variable for "use of health care" in an analysis explaining child health could result in overestimating the benefit of longer birth spacing. Summary The available evidence to date suggests that there is an important relationship between close birth spacing and poor child health. This association has been observed in a large number of diverse populations, both in developing countries and in high-mortality historical populations as well as in contemporary industrial- ized countries (Miller, 1989~. A substantial relationship persists even when controlling for several of the potentially confounding factors in this association. However, relatively little evidence is available about the physiological or behav- ioral mechanisms linking short birth spacing and child health or about confound- ing socioeconomic factors. Considerably more research is needed before we can draw definitive conclusions about the reasons for or causal nature of this associa- . tion or its magnitude. EFFECTS OF UNWANTED PREGNANCY The potential risks to a child's health of being born as the result of an unwanted pregnancy may be large. However, there is little direct evidence available on the subject, because distinguishing an unwanted from a wanted birth requires the measurement of attitudinal information about a couple's preferences and plans prior to conception. This information is sometimes not collected, and, when it is, it is usually measured after the birth, making its validity questionable. Tabulations from survey data, using retrospective reports of the wontedness status of births, do not show a consistent association between unwanted pregnancies and higher mortality risks. However, there is limited evidence from other sources suggesting that children born as a result of unwanted pregnancies are likely to experience greater health and psychological problems. Scrimshaw (1978) reviews anthropological re- search and suggests that parents are less likely or less able to take adequate care of children born as a result of unwanted pregnancies. There is evidence of the possibility of selective parental neglect from South Asia, where there is a strong preference for male children. The results of several studies (Simmons et al.,

OCR for page 53
66 CONTRA CEPrION ED REPRODUCTION 1982; Das Gupta, 1987; D'Souza and Chen, 1980; Bairagi, 1986; Chen et al., 1981) indicate that the higher mortality rates experienced by girls are due to poorer nutritional status and poorer care for girls who become ill. Weller et al. (1987) use survey data from the United States to show that women are not as likely to take adequate care of themselves during an unplanned pregnancy com- pared with a planned pregnancy. A study in Czechoslovakia by David et al. (1988) indicates that children whose mothers were denied requests for abortion suffer from a significantly higher incidence of psychological and developmental problems than other children. It is difficult to determine how applicable the findings of these last two studies are to families in contemporary developing countries. Nonetheless, this evidence suggests that children born as a result of unwanted pregnancies may be at higher risk. Grossman and Jacobowitz's results (1981) suggest that legalization of abortion in the United States may have contrib- uted to the decline in infant mortality by reducing the incidence of unwanted pregnancies. EFFECTS OF MATERNAL HEALTH CONDITIONS Although the fetus is well protected from most infections, there are maternal infections, mostly viral, and other conditions that can affect the fetus. The effect of maternal viral infection on infant and child health is a serious concern with the spread of HIV, especially among populations in sub-Saharan Africa, where the disease affects many women and children. Other maternal behaviors also put infants at greater risk, specifically smoking, drug use, and alcohol abuse. A recent report of the National Research Council (Turner et al., 1989) esti- mates the probability of HIV transmission from mother to infant is in the range of 30 to 50 percent. While considerable work on perinatal transmission of HIV remains to be done, the report also notes that some studies suggest that the risk of transmission is higher for infants born to mothers showing symptoms of HIV infection during pregnancy and those showing evidence of immunosuppression. HIV can also be transmitted from mother to infant via breast milk (Weinbreck et al., 1988~. Other important viral agents that may be passed from mother to fetus are toxoplasmosis, cytomegalovirus (CMV), rubella, hepatitis B virus, and herpes simplex. CMV and herpes simplex virus have been associated with fetal death, prematurity, intrauterine growth retardation, malformations, congenital infection, acute postnatal infection, and persistent postnatal infection. Rubella is associated with all of these except acute postnatal infection. Hepatitis B virus, which is endemic in Southeast Asia and other developing countries, has been linked to prematurity and fetal and neonatal infectious diseases (Overall, 1987~. Exposure to risks of sexually transmitted diseases and other genitourinary infections may be more likely among young women. Efiong and Banjoko (1975) found syphilis in 7 of 95 women in their first pregnancies younger than 16 in

OCR for page 53
REPRODUCE PACERS kD CRI=REN'S REALTH 67 urban Nigeria, compared with none in 100 older women in their first pregnancies selected as controls. A prospective study in Sierra Leone found that pregnant women under age 20 were more likely to have both urinary and genital tract infections in pregnancy than were older pregnant women (World Health Organi- zation, 1981~. Nonviral maternal infections are less likely to cross the placenta, but they may still affect the fetus before or during labor, especially if the membranes have ruptured prematurely. Infants born after premature rupture of the membranes are at increased risk of neonatal infections and respiratory distress syndrome. Matemal parasite, fungal, and bacterial infections that can affect the fetus include malaria, syphilis, and tuberculosis. In general, congenital infections may produce symptoms at birth, but in the majority of cases they first produce symptoms after some months. Even a congenital infection that is not itself a direct cause of infant deaths may leave the infant more susceptible to later infection. MINIMIZING THE RISKS OF CHILD DEATH As discussed above, assessing the potential impact of changes in reproductive patterns for child survival is complicated because there is not yet sufficient evidence to determine how much of the observed association is actually causal. Furthermore, the potential impact of fertility changes is likely to differ signifi- cantly, depending on whether one is considering the impact on individual chil- dren, on individual families, or on the mortality experience of the population as a whole. Although it should be clear by now that more information is needed to determine exactly how (and how much) changes in reproduction may affect child health, in this section and the next we attempt to provide estimates of the possible size of the effects of reproductive change on individual children and families in developing countries, based on currently available information. Risks to Individual Children Our estimates are based on the Hobcraft (1987) analysis of 18 developing countries and reflect averages across these countries. These estimates are in- tended to illustrate the potential implications of the Hobcraft results and those of similar analyses. If one assumes that the relationships between child mortality and reproductive variables observed in the Hobcraft analysis are causal, our calculations indicate the actual reductions in mortality risks for children and families that could be brought about by changes in reproductive patterns. In fact, it is likely that these figures are overestimates of the true causal effect. First, we use data from the 18 countries in Hobcraft's (1987) study to estimate an average child's probability of survival if he is born to a mother with a "better" reproductive history compared with an average child who was born to a mother I

OCR for page 53
68 CO=~CE=ION ID REPRODUCTION with a "worse" reproductive history. It is important to keep in mind when considering these results that many less-than-optimal reproductive patterns (such as very close child spacing) are not common in many countries, as we discuss further in Chapter 6. The estimates described here apply to risks for individual children, not to population-level mortality rates, which are affected by the distri- bution of maternal age, birth order, and birth spacing In the population. The issue of effects on population-level mortality rates is discussed in Chapter 6 and 7. In Table 5.2 we present estimated mortality rates from birth~~to age 2 for children born to women with "better" and "worse" spacing pattems, averaged across IS developing countries separately for teenage and older mothers. We have defined the reproductive pattern that is better for a child as having no birth either in the two years before or in the two years after his own birth. The "poor" spacing pattern is defined as having one both in the two years before the index child's birth and one birth in the two years afterward. Since all women who have children must, of course, have a first birth, we present these tabulations for children who are second or higher-order births. For purposes of companson, we also assume that the previously born child for both categories of reproductive histories has survived to the index child's birth. Children born to mothers ages 20 to 34 who have a better spacing pattern have a mortality rate of 67 deaths per 1,000 live births for 0- to 2-year-olds, which is almost half the rate for children born to women of the same age with a poor spacing pattern, and about 41 percent of the rate for children born to teenage mothers with a poor spacing pattern. It is clear from these figures, that, if the observed associations are causal, the potential gains to parents from improving spacing patterns and delaying childbearing until their twenties may be quite large, in terms of maximizing the survival chances for each of their children. Risks to Families Finally, we consider the risk to individual families of having a child die, given different mortality levels and different reproductive pattems. These results, shown in Table 5.3, are based on the same analysis (Hobcraft, 1987) discussed for Table 5.2 and are subject to the same caveats about causality and interpretation. It is important to note that the figures in Table 5.3 are from a simulation based on several assumptions and reflect average risks across results from 18 develop- ing countnes. Thus, they do not reflect the experience of families in any country. Like the figures in Table 5.2, their purpose is illustrative. The estimates show, for a group of 100 families, how many of their children would die before they reached their fifth birthday. The average number is obviously affected by the prevailing child mortality rate, and we have therefore carried out the simulation for three arbitrary baseline levels of mortality: SO, 100, and lSO deaths per 1,000 live births. The baseline levels of mortality reflect the risk of dying between both and age S for children with the lowest risk in the study population, i.e., children

OCR for page 53
REPRODUCTIVE PATTERNS AND CHILDREN'S HEALTH 69 TABLE 5.2 Estimated Average Mod ity Rates for Second- and Higher-Order Children to Women With Different Reproductive Pattems Better Spacing Pattem Poor Spacing Pattem Teenage Mothers 92 Mothers Ages 20 to 34 67 165 120 Note: "Better'' spacing pattem means that there were no births either in the 24 months before the index child or in the two years subsequent to the Tilde birth. "Poor" spacing pattem means that there was at least one birth during the 24 months before the index child and one birth in the two years after the index child's birth. For this comparison, in both cases, we assume the previous child survived. Source: Hobcraft (1987:Table 13). TABLE 53 Estimated Average Number of Child Deaths Expenenced by Families Under Different Conditions, Per 100 Families Baseline Child Mortality Rate Number of 50/1,000 Deaths 1OO/l,OOO Deaths lSO/l,OOO Deaths Children Closely Well Closely Well Closely Well Ever Bom Spaced Spaced Spaced Spaced Spaced Spaced 4-child family 45 23 92 45 142 68 6-child family 70 33 144 65 222 98 9-child family 112 50 232 100 358 150 Source: Calculations based An figures in Hobcraft (1987:Table 1). who are born at orders 2 or 3, who are well-spaced, and whose older siblings survived. We show simulations for families with 4, 6, and 9 children, because families who have more children are obviously going to be at greater risk of experiencing a child death, simply because they have more children. Estimates are shown separately for families in which all children are poorly spaced (birth intervals are all less than two years in length) and in which all children are well spaced (birth intervals are greater than two years in length).

OCR for page 53
70 CONTRACEPTION ED PRODUCTION The figures in Table 5.3 indicate that if the associations between reproductive variables and child survival are causal, families who space their children well are likely ~ loose fewer children than families who do not. For example, for a six- child family, the average number of child deaths experienced by families with well-spaced births is roughly half that for families with poorly spaced births. APPENDIX TABLE 5.A Studies of Infant and Child Health Location and Dependent Study Type of Data Variable Type of Analysis, Controls Aaby, Bukh, Guinea-Bissau; Incidence and Tabulations by HHC, A, CC, Lisse, and census and ease fatality; NS,E, SES Smits, 1984 health data rate of measles Bean, United States; Infant modality Digit regression; controls: Mineau, and nineteenth- PBI, PCS,hIA, Anderton, 1987 century Monnons, MAM, PD, PSBI, BO, R population -based longitudinal data Bijur, Great Britain; Accident Logit regression; controls: Golding, and longitudinal frequency HHC, SES,MMI, CHC Kurzon, 1988 data Boenna and van Kenya; Mortality during Log-linear regression; Vianen, 1984 population-based first week, controls: PBI, BO,MA. longitudinal data after the first Lifetable by SBI and week to 11 tabulations of BW and months, 12-23 mean weight and height months; binh- by PBI and SBI weight; mean weight and height at selected ages Cantrelle and Senegal; Duration of Tabulations by SSC, S. BO, Lendon, 1971 longitudinal breastfeeding; M, MB, BE, PCS, PSBI, PBI data infant and *lild mortality; fertility Chen, Hug, and Bangladesh; Mortality during Tabulations by A and S D'Sooza, 1981 population-based the first month, longitudinal data 1-11 months, 14 years, 5-14 years, 1544 years, 45- 64 years, and 65+ years; nutritional status; morbidity; diarrhea treatment

OCR for page 53
REPRODUCE PA"E~S ID CHI=REN'S HEALTH 7 ~ APPENDIX TABLE S.A Continued Location and Dependent Study Type of Data Vanable Type of Analysis, Controls Clark, 1981 Guatemala; Infant growth OLS regression; population-based (change in controls: SES, B-F, SP, R. longitudinal weight between MW, MH, BO, MA, SB, PBI, data, cross- birth and so HHC, S. BW sectional socio- months, six and economic survey twelve months, birth and twelve months) Cleland and Pakistan; Mortality during Log-linear regression; Sathar, 1984 WF~retrospective first month, 1-11 controls: PBI, PCS,SES, fertility history months, 12-23 R. S. BO, MA, PSBI, PI, BF, months, 24-59 SBI months. Costello, Uganda; Nutritional OLS Regression; controls: 1986 Household survey, status MA, MP, ML, PP, PL, P. L retrospective ME, MS, MAM, MG, fertility history, R. SES medical data DaVanzo, Malaysia; Mortality during OLS and logit regression; Butz, and Malaysian family first week, 8-28 controls: MA,PSBI,SB, Habicht, 1983 life survey- days, 2~ months, PBI, S. BW, BO, BF, SES, YB, retrospective 7-11 months, ~11 HHC, D, E, R fertility history months DaVanzo, Malaysia; Birthweight Logit, OLS, and variance- Habicht, and Malaysian family components least squares Butz, 1984 life survey- regression; controls: Retrospective S. FB, MA, AM, PSBI, PBI, fertility history SES, ON, R. E, YB Doyle, Morley Nigeria; Birth interval; Tabulations by BO, PCS, Woodland, and population-based birthweight; PCCS, SCS, PBI, SBI, K Cole, 1978 longitudinal data mean growth D'Sooza and Bangladesh; Infant and child Tabulations by YB, S. Chen, 1980 population-based mortality; MB, CD longitudinal data mortality during 5-14, 15-44, 45- 64 and 65+ years Fedrick, and Great Bntain; Stillbirths; Tabulations by PBI,SES, Adelstein, cross-sectional neonatal MA, CD,PCS 1973 perinatal mortality; mortality survey birthweight

OCR for page 53
72 CONTRACEPTION ED PRODUCTION APPENDIX TABLE 5.A Continued Location and Dependent Study Type of Data Variable Type of Analysis, Controls Fleming and India; Risk of mat- Logic regression; Gray, 1988a Narangwal nutrition controls: SBI, S. C, NS Nutrition and at selected Health Intervention ages. Project, longitudinal data neming and India; Birthweight; Logit regression; Gray, 1988b Narangwal Nutrition infant end controls: S,C, PBI, BO, and Health later- child growth. PCS vexation Project, longitudinal data Fortney and Iran; Infant modality Digit regression; Higgins, 1984 hospital-based before the controls: PBI,BO,MA data mother's dis- charge from the hospital; binhweight Gubhaju, 1986 Nepal; Infant and Logit regression; WFSretrospective child mortality controls: BO, MA, S. PBI, fertility history PCS, SES, R. YB Hobcraft, 1987 34 countries; Mortality during Log-linear regression; WFS retrospective first month, controls: PBI, SBI, BO, fertility histories 1-11 months, MA,S,SES; 12-23 months, tabulations by family 24-59 months, fonnation patterns 0-4 years (classification based on MA, BO, PBI, SBI) Hobcraft, 39 countries; Mortality during Log-linear regression; McDonald, and WFSretrospective first month, controls: PBI, SBI, BO, Rutstein, 1985a fertility histories 1-11 months, MA, S. SES 12-23 maths 24-59 months Knodel and Germany; Mortality during Multiple classification Hermalin, 1984 nineteenth-century first month, analysis; German villages, 1-11 months, tabulations by MA, BO, SS, populaiion-based 0-11 months, PBI, PCS; longitudinal data 12-59 months controls: MA, PBI, R. PD

OCR for page 53
REPRODUCTIVE PATTERNS AND CHILDREN'S HEALTH 73 APPENDIX TABLE S.A Continued Location and Dependent Study Type of Data Variable Type of Analysis, Controls Koenig, Bangladesh; Mortality during Hazard model; controls: Phillips, population-based first month, S. MA, BO, SES,-PBI, SBI Campbell, and longitudinal data 1-11 months, D'Souza, 1988 12-23 months, 24-59 months Palloni and 12 Latin American Mortality during Logit regression; Millman, 1986 countnes; 1-2 months, 3-5 controls: SES,MA,R, S. WFSretrospective months, ~11 MB, SB, BO, BF, PBI, SBI fertility histories months, 12-59 months Pebley, Knodel, Germany; Infant modality Logit regression; and Hermalin, nineteenth-century by birth rank controls: PCS, PBI,MA, 1988 German villages, SES, R. PD population -based longitudinal data Pebley and Guatemala; Infant and Hazard model; Stupp, 1987 femalelifo history child mortality controls: MA, BO, S. D, survey, cross- PCS, SES, YB, PBI, SBI, BF . . sectlona. somo- econom~c survey Rosenzweig and United States; Birthweight Two-stage least squares Schultz, 1983 national nasality regression; followback surveys, controls: first stage- local area price, SES, R. HE, HDFP, PR, HFP, health, and labor DP, J. E, UK, HB, ST; force data second stage DO, SM, PAR, MA,E Rutstein, 1983 41 countries; Mortality during Tabulations by YB, S,MA, WFS- retrospective first month, BO, PBI,PCS,MB birth histories 1-11 months, 0-11 months, 12-23 months, 24-59 months Weller, United States; Pregnancy Logit regression; Eberstein, and cross-sectional wontedness controls: E, SES,R, BO Bailey, 1987 nasality survey measured by two indicators- cigarette smoking and timing of prenatal care

OCR for page 53
74 CONTRACEPTION AND REPRODUCTION APPENDIX TABLE 5.A Continued Study Location and I dependent Type of Data Variable Type of Analysis, Controls Wolfers and Ecuador, Scrimshaw, 1975 etrospective fertility and sexual union histories Birth interval length; pregnancy outcome; mortality during the first month, 1-11 months, 0-11 months, 12-23 months, 24-59 months Tabulations by MA, BO, YB, PBI, SBI, PCS, PSBI Note: Hobcraft et. al. (1985) presents regression estimates for 35 of the 39 countries included in the discussion. Hobcraft (1987) reviews findings for 34 of the 35 countries included in the regression analyses of Hobcraft et. al. (1985) and presents new analyses for 18 of the 34 countries by family formation pattems. Key to Abbreviations AM BE BO BW C CC CD CHC D DD DN DP E FB HB HDFP HE HFP HHC age group mother's age at menarche breastfeeding birth or pregnancy order birth weight caste clustering of cases cause of death child characteristics, such as aggression, overactivity, independence type of delivery number of months pregnant before mother consulted a doctor or nurse distance to nurse number of doctors and OB-GYNs per capita ethnicity first birth hospital beds per capita number of health departments with family planning services per capita local government health and hospital expenditures number of hospitals with family planning services per capita household composition percentage of persons employed in manufacturing, service, or government jobs history of kwashiorkor currently breastfeeding month of birth maternal age M MA MAM mother's age at marriage MB multiple births ME menstruating MG presence of malaria or gonorrhea MH mother's height

OCR for page 53
REPRODUCTIVE PATTERNS AND CIlI=REN'S IlEALTH 75 ML months breastfeeding MM:I maternal malaise inventory, a measure of the psychological well-being of the mother MP months pregnant MS marital status MW mother's weight NS nutritional status currently pregnant PAR number of live births bom to mother PBI previous birth or pregnancy interval PCCS survival of a child prior to the preceding child PCS survival of previous child PI prior interval (birth interval immediately prior to the preceding interval) PD previous infant deaths PL months breastfeeding/months exposed to conception PP months pregnantimonths exposed to conception PR cigarette and mills prices PSBI R S SB SBI SCS SES SP SM SS SSC ST UE YB proportion of other pregnancy intervals that are short or average of birth intervals urban/rural residence or region sex of child proportion or number of stillbirths subsequent birch interval survival of subsequent child education, income, occupation, housing characteristics supplementary food number of cigarettes smoked per day by mother while pregnant sibship size survival status of the index child sales tax on cigarettes general and female unemployment rate year of birth